Abstract
Background: In recent decades, protein-protein interaction (PPI) prediction has become a popular research area and many classifiers have been developed for PPI prediction. However, no single classifier has been able to predict PPI with high-confidence. We postulate that combining individual classifiers could improve PPI prediction accuracy over individual classifiers.
Methods: We developed protein-protein interaction classifiers merger (PPCM) that combines output from two PPI prediction tools -- GO2PPI and Phyloprof using random forests algorithm. The performance of PPCM was tested by area under curve (AUC) using a gold standard PPI database that contained positive PPI pairs as well as negative PPI pairs.
Result: Our AUC test showed that PPCM significantly improved PPI prediction accuracy over corresponding individual classifiers. We found that more classifiers incorporated into PPCM led to better improvement in the PPI prediction accuracy. Furthermore, cross species PPCM achieved competitive and even better prediction accuracy compared to the single species PPCM.
Conclusion: This study established a robust pipeline for PPI prediction by integrating multiple classifiers using random forests algorithm. This pipeline could be very useful for predicting PPI in non-model species.
